musing ai monads

15/09/2011: this AI book (Artificial Intelligence: A Modern Approach) is good so far. hopefully it's all like this, not just ch1 ;) am thinking a test to help with archive project would be good. [1]

prob more of a parser than AI but maybe useful to learn some natural language processing & machine learning & automated reasoning #terms [2]

saw mention of using money as the 'reward' for the atoms/monads to learn. thinking of using music instead. like real life in that community [3]

music taste -> street cred -> popularity(?) of sorts. have to know history/share knowledge/durations of participation etc. [4]

I wonder if you could model the "real" people in the virtual system (memories) & see if they're cred matches what happen[ed|s] real lift too [5]

ideas form at 11pm after musing. by morning they no longer seem feasible ha [6]

in the video, Ben Goertzel mentions that it's often hard to work out how well the AI agents / programs are working - to measure your successes. maybe this could be a simple test - to model the people / artists virtually based on information about them from other people's memories, or articles / photos / flyers etc mentioning them. could you end up with virtual people with the same level of "street cred" as in a real "underground" / grassroots music community. apply different weightings according to different sources - of course the djs / bands would get more promo type documentation, but others in the scene would get word-of-mouth cred also. need to adjust these, so it more closely models the real world. like an "artificial life" system / simulation, but based on information / documentation of real people.

the weighting of the documentation / information / "mentions" would have to be calculated so that "good" / worthwhile / reliable sources - ala wikipedia style, but perhaps not as strict - ie using blogs and music related sources are ok, since they are usually by passionate people who curate / filter the info / listings. just pure promo spam, has a negative count, to discourage it. also, of course would have to take into account the person's contribution to the community - not just the ones who are good at promo etc, though they tend to float to the top in popularity stakes - not necessarily the only ones in street cred / good music taste stakes though. but then there's the question of what's "good" - but people in the community know (same for any community really) - might not be as obvious to people less familiar, but is "common ground" of sorts with people within the community. maybe it'd turn out like a machine "people watching" app / scenario but help document / archive the period & community at the same time. could the system (eventually) reply with different responses - eg like the Philip K Dick / Einstein robots (robot heads) which could have a conversation. &/or maybe a MOO like the old days - text based to start with? a twitterMOO? (though there was no twitter when this all happened) (musing out of control here now)

hrmm... could it work. at least it might help with collating and sorting the information if I can at least get that part working, then add some basic weightings and a final score. and compare. why do this at all? would it matter? probably not, but as a measurement exercise it would be interesting too perhaps.

ai (artificial intelligence) class & ml (machine learning) class & db (database) class are coming up soon - starting in october. not sure if we can do our own projects or if there'll just be set exercises, but if so, then this might be a nice little project to work on. there's always after class too. (if I can handle the maths - has been a while since I've done that kind of maths)

collated to: (Brisbane archive) (Australian archive)
using drupal - CCK & views...
another custom database & code
servlet / XML / JSON ?

also list / link to some other ephemera around at that time to help set the mood & trigger memories / the vibe. influential books / music / videos / ideas etc (ala mondo 2000 / cyberpunk / early tech / moos / BBS / fashion etc )


16/09/2011 musings
section 1.3 triggers thoughts

ha. love it that AI couldn't translate "the spirit is willing but the flesh is weak" it became "the vodka is good but the meat is rotten" [7]

maybe the AI needs to gain some spirituality to become closer to humans? can it even do this when it knows who it's maker was [8]

it's kinda nice not knowing. opens the imagination to possibilities. computer just reads a history of the computer/AI 101 & gets depressed [9]

maybe that's how the Mind & consciousness day there was a random thought then they grew to help consider / muse / dissect it all [10]

"expert systems" there's a term I haven't heard for a while. since 90s computing if I remember correctly. must dig out some old mags [11]

could you make an AI that learnt via looking at and storing pictures. I think this is how my brain works at least. it's what I see in my inner mind when I need to remember something - see the page it was written on, or the place something was left, or the scenario of the event / problem (if a work issue, when remembering how it was fixed previously). could you use flickr or youtube or soundcloud or similar systems to outsource the memory (like I do) to these places for filestorage. maybe remember the tags / keywords. and search and retrieve. I know there's a lot of maths in AI research, and I suppose my brain is doing some maths, but it's just not obvious to me that it's happening except for a specific maths/athremetic problem (maybe just in my head ;) I know it's also hard for computers to view images too. but I also know there's lots of image processing code out there - with the maths done already - so maybe this could be used for the archive project for looking at the flyers. of course it wouldn't probably be able to read the text on them, unless there's some sort of OCR in the system? or metadata / text as part of the cataloguing. wouldn't it be nice if the AI (would-be AI) could do all the cataloging! save time!

--- finished chapter 1!
nice listening music for reading - "infected by radio" <3

::: location: